The A_2A-adenosine receptor: a GPCR with unique features?

Abstract

The A_2A-adenosine receptor is a prototypical G_s-coupled receptor. However, the A_2A-receptor has several structural and functional characteristics that make it unique. In contrast to the classical model of collision coupling described for the β-adrenergic receptors, the A_2A-receptor couples to adenylyl cyclase by restricted collision coupling and forms a tight complex with G_s. The mechanistic basis for this is not clear; restricted collision coupling may arise from the interaction of the receptor with additional proteins or due to the fact that G protein-coupling is confined to specialized membrane microdomains. The A_2A-receptor has a long C-terminus (of >120 residues), which is for the most part dispensable for coupling to G_s. It was originally viewed as the docking site for kinases and the β-arrestin family to initiate receptor desensitization and endocytosis. The A_2A-receptor is, however, fairly resistant to agonist-induced internalization. Recently, the C-terminus has also been appreciated as a binding site for several additional ‘accessory' proteins. Established interaction partners include α-actinin, ARNO, USP4 and translin-associated protein-X. In addition, the A_2A-receptor has also been reported to form a heteromeric complex with the D₂-dopamine receptor and the metabotropic glutamate receptor-5. It is clear that (i) this list cannot be exhaustive and (ii) that all these proteins cannot bind simultaneously to the receptor. There must be rules of engagement, which allow the receptor to elicit different biological responses, which depend on the cellular context and the nature of the concomitant signal(s). Thus, the receptor may function as a coincidence detector and a signal integrator.

Adenosine receptor classification

G-protein-coupled receptors (GPCRs) for adenosine, which either inhibited (A₁) or stimulated (A₂) cAMP accumulation, were first identified in the 1970s (van Calker et al., 1979) and differentiated—by structure–activity relationship and other pharmacological criteria—from the direct action of adenosine at the catalytic centre of AC, the so-called P-site (Londos et al., 1980). The further subclassification of A₂-adenosine receptors was introduced later to account for the receptor that required low concentrations of adenosine and adenosine analogues for the stimulation of AC; this was termed the A_2A receptor (Bruns et al., 1986). The A_2A receptor has one additional close relative the A_2B receptor, which also signals via G_s, and two more distant relatives termed A₁- and A₃-adenosine receptors, which interact with pertussis-toxin-sensitive G proteins of the G_i and G_o family. These receptors are encoded by distinct genes and are classified based on their affinities for adenosine analogues and methylxanthine antagonists (Klinger et al., 2002a).

Interest in the A_2A receptor has been rekindled by findings that suggest that both agonism at this receptor and its inhibition by specific antagonists may be of therapeutic relevance. Examples are highlighted in the accompanying reviews (Brown, 2008; Sitkovsky, 2008). The A_2A-adenosine receptor, however, is also of interest, because it has many features that do not readily fit into the accepted paradigms of GPCR signalling and desensitization; in addition, the receptor binds many proteins other than G_s. This review will focus on these three points.

Precoupling, tight coupling and restricted collision coupling

In contrast to β-adrenergic receptors (or rhodopsin), which engage their signalling cascade by collision coupling, the A_2A receptor has long been known to couple to AC by restricted collision coupling. Restricted collision coupling refers to the inability of the receptor to access all G_s molecules and thus activate the full complement of all AC moieties in the cell membrane. Originally, these observations were made with the avian A₂ receptor in turkey erythrocytes (Tolkovsky and Levitzki, 1978; Braun and Levitzki, 1979). Restricted collision coupling, however, was also observed in human platelets, which express A_2A receptors (Gross and Lohse, 1991; Lohse et al., 1991). The A_2A receptor forms a tight complex with G_s; this complex is resistant to dissociation by guanine nucleotides. In addition, the receptor can be solubilized in a complex with G_s in the absence of agonist (Nanoff et al., 1991; Nanoff and Stiles, 1993). Thus, the A_2A receptor appears to be precoupled to G_s. Precoupling, that is, formation of a stable complex between receptor and G protein, by definition, gives rise to restricted collision coupling.

It is also worth noting that, when measured directly by fluorescence/Foerster resonance energy transfer (FRET), the kinetics of G protein activation do not differ substantially between β₁-adrenergic receptor and A_2A-adenosine receptor (Hein et al., 2006). These experiments, however, cannot decide on the existence or absence of precoupling. First, in the ground state (that is, in the absence of agonist), A_2A receptor and G_s may be in a complex that does not support efficient FRET. Agonist binding reorients both receptor and G_s into a tightly packed complex, in which energy transfer is supported. Second, when compared to the time constants required for G protein activation (and formation of the high-affinity complex of agonist, receptor and G protein), diffusion of the receptor does not appear to be rate limiting; ternary complexes formed by fusion proteins of receptor and α-subunits of G protein are kinetically indistinguishable from those formed by the individual components (Waldhoer et al., 1999). Although precoupling of the A_2A receptor to G_s may provide a mechanistic basis for restricted collision coupling, it remains unclear as to why the A_2A receptor would form a tight complex with G_s. Earlier reconstitution experiments did not suggest that the A_2A receptor (Nanoff et al., 1994) had a substantially higher affinity for G_s than the β₁- or β₂-adrenergic receptors (Freissmuth et al., 1991). The formation of a tight receptor/G_s complex and the restricted mobility of the A_2A receptor may, however, arise from the interaction with additional proteins (see below); the receptor may, for instance, be concentrated in membrane subdomains, because it is tethered to the actin cytoskeleton via its binding partner α-actinin (Burgueño et al., 2003). Alternatively, the interaction between A_2A-adenosine receptor and G_s may be limited to membrane microdomains (‘lipid rafts'). We have examined these conjectures and have observed that the actin cytoskeleton did not account for the restricted mobility of the A_2A receptor but that coupling of the A_2A receptor to G_s was contingent on the presence of cholesterol. In contrast, activation of an alternative signalling pathway (namely ARNO (ARF-nucleotide binding site opener), the guanine nucleotide exchange factor for ADP-ribosylation factor (ARF)6; Gsandtner et al., 2005) was not impaired by removal of cholesterol (Charalambous et al., manuscript submitted).

The other issue is why should have receptors evolved to differ in their coupling modes, that is, collision coupling versus restricted collision coupling. This question could simply be considered an academic preoccupation of little relevance. However, it is evident that collision coupling and restricted collision coupling confer different outcomes to adaptive changes; if a receptor operates by collision coupling, increasing the number of receptors primarily augments the sensitivity of the system. In other words, the concentration–response curve for agonists is shifted to the left, because all receptors have unrestricted access to all G proteins. In restricted collision coupling, where the receptor is allocated its particular lot of G proteins, there are no spare receptors. Changes in receptor number, therefore, translate in changes of the maximum response E_max rather than in EC₅₀ values. Actually, this is precisely the finding that was obtained when the number of A_2A receptors was decreased by inactivation (Lohse et al., 1991) or raised by facilitating its export from the endoplasmic reticulum (Milojevic et al., 2006).

Desensitization and internalization of the A_2A receptor

For GPCRs, desensitization and receptor trafficking is important for regulation of the temporal and spatial aspects of their signalling properties. Both desensitization and internalization function to control signal termination and rely on the intracellular binding of GPCR kinases (GRKs) and arrestins. The basic mechanisms underlying desensitization of GPCRs is understood in considerable detail. The major insights have originally been obtained with two model GPCRs, rhodopsin and the β₂-adrenergic receptor; the agonist stabilizes the active conformation of the receptor, which is recognized by GRKs (GRK2–6, with GRK1 being the specialized rhodopsin kinase). Phosphorylation by GRKs of clustered serine/threonine residues creates docking sites that allow the recruitment of arrestins. The general arrestins, arrestin 2 and arrestin 3, are also referred to as β-arrestin 1 and β-arrestin 2, whereas cone and rod arrestins are confined to the visual system. Binding of arrestin 2 and arrestin 3 via its N-terminal phosphate sensor triggers a structural rearrangement that exposes their C terminus, which recruits components of the clathrin coat. Accordingly, binding of arrestins favours endocytosis of receptors via the clathrin-dependent pathway (reviewed by Gurevich and Gurevich, 2006). It is a matter of debate whether the A_2A receptor undergoes a desensitization/resensitization cycle under physiological conditions; short-term desensitization can be observed in various cell culture models (reviewed by Olah and Stiles, 2000). The structural requirements of the receptor that may be necessary for the rapid desensitization have been examined in receptor mutagenesis studies in chinese hamster ovary cells (Palmer and Stiles, 1997). The removal of 95 amino acids, including 10 potential phosphorylation sites, from the cytoplasmic tail of the A_2A receptor had no effect on the ability of the receptor to undergo rapid desensitization. Although a simultaneous mutation of Thr298 and Ser305 to Ala residues attenuated the desensitization observed in response to short-term agonist treatment, it did not block the ability to desensitize after long-term agonist exposure. Analysis of the individual mutation of these residues revealed that mutation of Thr298 alone was sufficient to diminish both short-term desensitization and agonist-stimulated receptor phosphorylation. This is remarkable because efficient recruitment of arrestins typically requires a cluster of phosphates, which interact with the N-terminal phosphate sensor and thereby trigger the structural rearrangement that leads to the tight interaction between arrestin and receptor (Gurevich and Gurevich, 2006). This issue may be explained by the following two alternative scenarios: (i) negatively charged amino acids in the vicinity of Thr298 may substitute for phosphoserines/phosphothreonines and thus obviate the requirement for multiple phosphorylated residues for arrestin recruitment or (ii) phosphorylation of Thr298 may induce a conformational change, which, per se, suffices to disrupt G protein coupling; in this scenario, recruitment of arrestin would be dispensable. A precedent for this latter model is provided by the observation that PK-A-mediated desensitization of the β₂-adrenergic receptor is apparently not dependent on arrestin recruitment (Vaughan et al., 2006).

Manipulations in the level of GRK2 affect the extent of desensitization of the A_2A receptor (Mundell et al., 1997, 1998). However, the relevance of these observations has been questioned for the following reasons: (i) adenosine appears to signal tonically via the A_2A receptor; in vivo, many actions of adenosine are accounted for by the retaliatory metabolite concept that hypoxia and tissue damage lead to the accumulation and release of adenosine; adenosine acts via its receptors to orchestrate a concerted protective response, where the A_2A receptor plays a prominent role (Linden, 2005). From a teleological perspective, rapid desensitization is of questionable value in a situation where sustained signalling is vital to support long-term tissue repair. (ii) In most instances, it has been difficult to visualize internalization in vivo (Fredholm et al., 2001). A representative experiment is shown in Figure 1, where the yellow fluorescent protein-labelled A_2A receptor was visualized together with cyan fluorescent protein-tagged Rab5 (a marker for early endocytotic vesicles). Even continuous administration of agonist did not result in any appreciable increase in basal A_2A receptor endocytosis. There are, however, reports that suggest that rapid internalization upon agonist binding can occur via an interaction with the actin-binding protein α-actinin (Burgueño et al., 2003). α-Actinin is a component of the actin cytoskeleton that plays a central role by directly crosslinking actin molecules; so the presence of a complex involving A_2A receptor and α-actinin suggests that α-actinin may mediate receptor association with the actin cytoskeleton. The precise regulatory role of this interaction remains to be established.

Figure 1.

Human embryonic kidney cells were transiently transfected with plasmids coding for the fluorescence-tagged versions of the A_2A receptor and Rab5. Cells were then stimulated with the A_2A-specific agonist CGS2168. Images were taken at the indicated times using a confocal laser microscope. For better visual distinction in the overlay, fluorescent A_2A receptors were converted to red and Rab5 to green.

Sustained G-protein-independent signalling via ARNO

The action of GRKs on agonist-liganded receptors and the subsequent binding of arrestins to the receptors are necessary to initiate desensitization and a second wave of long-term signalling events, the persistent stimulation of mitogen-activated protein kinase (MAP kinase) being a particularly well-studied example (Lefkowitz and Shenoy, 2005). The A_2A-adenosine receptor was first shown to cause a G_s-independent sustained stimulation of MAP kinase in human macrovascular endothelial cells (Sexl et al., 1997), which is presumably related to its capacity to support endothelial cell proliferation (Sexl et al., 1995). It was later appreciated that depending on the cell type studied, there was a G_s- and cAMP-dependent mechanism that supported MAP kinase activation, for example, in PC12 cells (Seidel et al., 1999) and that required src-like kinases (Klinger et al., 2002b), as well as a G_s-independent pathway that required p21^ras (Seidel et al., 1999). Heterotrimeric G proteins were, in fact, dispensable for the latter signal, as was dynamin-dependent internalization (Klinger et al., 2002c). Clearly, based on these findings, it is safe to conclude that the A_2A receptor does not conform to the paradigm established for the β₂-adrenergic receptor. More recently, the G_s-independent signalling pathway was found to comprise ARNO and ARF6 (schematically represented in Figure 2); ARNO/cytohesin-2 was identified as an interactor that bound to the C terminus of the A_2A receptor in an yeast two-hybrid screen (Gsandtner et al., 2005).

Figure 2.

Schematic illustration of the signalling pathways of the A_2A-adenosine receptor. The link between ARF6 and Ras remains to be elucidated.

ARNO contains a sec7 domain, which acts as the guanine nucleotide exchange factor for ARF-family members. These small G proteins have been recently recognized to play a prominent role in endocytosis of GPCRs; many but not all receptors require ARF6 (originally identified as the cofactor necessary for cholera-toxin-mediated ADP-ribosylation of Gα_s) for internalization (Lahuna et al., 2005; Houndolo et al., 2005). ARF6 differs from the other members of the ARF family because it is primarily found at the cell membrane, where it regulates endocytosis through both the clathrin/dynamin-dependent and a less well-understood clathrin/caveolae-independent pathway. Two mechanistically conflicting models have been proposed, both involving the exchange factor ARNO. During internalization of the β₂-adrenergic receptor, arrestin binding precedes that of ARNO; in fact, ARNO is recruited by arrestin (Claing et al., 2001; Lefkowitz and Shenoy, 2005). In contrast, for the luteinizing hormone/human chorionic gonadotropin receptor, arrestin binding is contingent on prior stimulation of ARNO, the ensuing activation of ARF6 required for the release of membrane-associated arrestin (Hunzicker-Dunn et al., 2002). The A_2A receptor differs from these two paradigmatic examples because it binds ARNO/cytohesin-2 directly (Gsandtner et al., 2005). Truncation experiments showed that the proximal 22 amino acids of the C terminus are sufficient for the interaction (highlighted in Figure 3). The classical pathway, activation of AC, was not altered by ARNO, and this was also true for agonist-induced desensitization. In contrast, expression of dominant-negative ARNO and dominant-negative ARF6 abrogated the sustained phase of MAP-kinase stimulation induced by the A_2A receptor. Thus, the binding of ARNO to the proximal portion of the C terminus provides one of the missing links to support the alternative—heterotrimeric G-protein-independent—signalling pathway of A_2A receptor.

Figure 3.

Scheme of interaction sites on the C terminus of the A_2A-adenosine receptor. Binding sites for α-actinin, ARNO, the D₂ receptor and USP4 have been tentatively assigned in Burgueño et al. (2003), Gsandtner et al. (2005), Woods and Ferré (2005) and Milojevic et al. (2006), respectively. The orange circle identifies a threonine important for short-term desensitization (Palmer and Stiles, 1997). The red circle depicts the arginine unique to the A_2A receptor. All other adenosine receptors have a cysteine in this position, which in most GPCRs is palmitoylated and thus stabilizes helix 8.

Accessory proteins

In the last 15 years, it has been appreciated that GPCRs bind many additional proteins; these are being referred to as ‘accessory proteins' for lack of a better description. The list includes proteins involved in selective retention in specialized membrane compartments (for example, dendritic spines) or components of signalling cascades (for example, PKs of the Src and Jak family, adapter proteins like Grb2 and Shc, and enzymes generating second messengers such as PLCβ1 and eNOS; for a detailed list, see Bockaert et al., 2004). The A_2A-adenosine receptor can also claim a respectable list of accessory proteins (Gsandtner and Freissmuth, 2006) that bind to various portions of its C terminus (Figure 3). In addition, A_2A receptors, like other GPCRs, can form a heteromeric complex with at least one GPCR, namely, the D₂-dopamine receptor (see below), and it can transactivate the neurotrophin receptors TrkA and TrkB (Lee and Chao, 2001). The term transactivation of TK has been coined by analogy to its original use in the description of gene regulation: GPCRs can act in trans, that is, recruit signalling components by activation of TK receptors in the absence of cognate ligands of the latter (in this instance, in the absence of nerve growth factor and related neurotrophins). The molecular basis of transactivation is poorly understood, and it is specifically not clear whether the A_2A receptor forms a heteromeric complex with TrkA and/or TrkB.

Interaction of the A_2A-adenosine receptor with other GPCRs (D₂-dopamine receptor, mGluR5)

The first indirect evidence for a membrane-delimited interaction between the A_2A receptor and the D₂-dopamine receptor came from binding experiments that suggested that activation of the A_2A receptor interfered with coupling of the D₂-dopamine receptor to its cognate G proteins, presumably a mixture of G_i and G_o isoforms (Ferré et al., 1991). Confocal laser microscopy showed that A_2A receptors and D₂-dopamine receptors colocalized to a large extent in the cell membranes of stably transfected neuroblastoma cells and in cultured striatal neurons; in addition, the existence of heteromeric complexes between these two receptors was confirmed in co-immunoprecipitation experiments (Hillion et al., 2002). Evidence for a direct and specific interaction between A_2A and D₂ receptors was also obtained with a quantitative bioluminescence resonance energy transfer analysis and sensitized emission FRET as well as acceptor photobleaching FRET analysis (Canals et al., 2003; Kamiya et al., 2003). This phenomenon may therefore constitute the molecular basis of the A_2A/D₂ receptor interactions, which give rise to mutual antagonism in vitro and in vivo (Fuxe et al., 2005, 2007). These findings led to the idea that blockade of the A_2A receptor could be of value in the treatment of Parkinson's disease. In fact, istradefylline, a selective A_2A receptor antagonist has shown (modest) efficacy in phase II trials (Hauser et al., 2003), which were considered promising enough to justify phase III clinical trials (Schapira et al., 2006); related selective antagonists are also being tested (Schapira et al., 2006). The metabotropic glutamate receptor 5 (mGluR5) has also been suggested to exist in a complex with the A_2A receptor in the striatopallidal neurons (Ferré et al., 2002). This is thought to explain the complex interaction of mGluR5 and A_2A receptor ligands; the beneficial action of an mGluR5 antagonist is enhanced by blockage of A_2A receptor, but it is nevertheless contingent on the presence of A_2A receptors (and D₂ receptors) (Kachroo et al., 2005). It has to be stressed that the interaction of D₂ and A_2A receptors does not uniformly result in mutual antagonism. Under appropriate conditions, the two receptors can also signal synergistically (Kudlacek et al., 2003), and this has been proposed to be important for enhanced rewarding action of adenosine in the nucleus accumbens (Yao et al., 2002). Finally, heterodimerization may affect the kinetics of desensitization; this has, for instance, been documented for the trafficking pattern of V_1a- and V₂-vasopressin receptors (Terrillon et al., 2004). In fact, the A_2A receptor has been shown to co-internalize with the D₂-dopamine receptor (Hillion et al., 2002), suggesting that the D₂-dopamine receptor confers a classical GRK/arrestin-dependent internalization mechanism to the A_2A-adenosine receptor.

Interactors that bind to the C terminus of the A_2A receptor

In contrast to many other GPCRs and, in particular, to the other members of the adenosine receptor family, the A_2A receptor has an unusually long intracellular C-terminal tail (122 amino acids in man compared with only 34 residues in the C terminus of the A₁-adenosine receptor). The juxtamembrane segment immediately adjacent to the seventh transmembrane helix is required for proper folding of the receptor. The rest of the C terminus (100 amino acids) is dispensable for ligand binding (Piersen et al., 1994) and for G protein coupling (Klinger et al., 2002c). The vast majority of group I (rhodopsin-like) GPCRs carry one (or two) palmitoylated cysteine(s) within the proximal portion of their C terminus (typically, approximately 20 residues away from the end of the seventh transmembrane helix). The palmitate thioester is thought to act as an additional anchor. This stabilizes the proximal segment in an α-helical conformation (referred to as helix 8 in the rhodopsin structure, which is oriented in a manner parallel to the membrane and perpendicular to helix 7). The A_2A receptor does not have any cysteine residue in the proximal segment; there is only a single cysteine in position 394 in the human receptor and this is absent in other species orthologues (for example, of rat and mouse). Instead of the canonical cysteine, the A_2A receptor carries an arginine residue in position 309 (circled in Figure 3). Thus, one is tempted to speculate that the C terminus of the A_2A receptor is more flexible because it is not constrained by a lipid anchor. These two features, relative length and flexibility, may combine to afford the interaction of the A_2A receptor with many additional proteins, other than G protein arrestins and kinases, that is, the so-called ‘accessory' proteins.

In recent years, several interaction partners were identified in yeast two-hybrid interaction screens using the C terminus of the A_2A-adenosine receptor as a bait and the list is growing rapidly. Figure 3 gives an overview of accessory proteins that have been found to bind to the A_2A receptor. Although the C terminus may exist in an extended conformation and thus provide a lot of room for interaction, the size of the individual binding partners makes it unlikely that the C terminus can accommodate several partners simultaneously. This is highlighted in Figure 3 by the overlapping binding sites for α-actinin (Burgueño et al., 2003) and ARNO (Gsandtner et al., 2005). Thus, individual interactions are likely to be only transient. The nature of the regulatory inputs that promote and terminate these interactions is unknown, but Thr298 (highlighted by an orange circle in Figure 3) is strategically placed to regulate binding of ARNO and α-actinin. It is therefore conceivable that the interaction of the receptor with ARNO and/or α-actinin is regulated by phosphorylation of Thr298. Similarly, the hetero-oligomer between D₂ and A_2A receptors has been proposed to be stabilized by binding of the phosphorylated Ser374 (boxed in orange Figure 3) to an arginine-rich segment in the third intracellular loop of the D₂ receptor (Woods and Ferré, 2005). However, the authors used only synthetic peptides to show the stability of the arginine–phosphate electrostatic interaction. It remains to be demonstrated that the phosphorylation of Ser374 of the A_2A receptor occurs in vivo and that it is a prerequisite for complex formation between A_2A and D₂ receptors.

Finally, it is also conceivable that ubiquitin-specific protease (USP)4 may also participate in regulating interactions. The binding site of deubiquitinating enzyme USP4 on the A_2A receptor has been located within the last 50 amino acids of the C terminus and USP4 is required for export of the receptor from the endoplasmic reticulum (Milojevic et al., 2006). It is, however, conceivable that USP4 may also regulate later events in the trafficking of the receptor. The endocytotic routing of GPCRs has been known to depend on the extent of their ubiquitination (Wojcikiewicz, 2004; Lefkowitz and Shenoy, 2005). Extensive ubiquitination favours trafficking of endosomes to late endosomes and eventual lysosomal degradation over recycling (Marchese and Benovic, 2001; Shenoy et al., 2001). It is evident that for recycling receptors both phosphorylation and ubiquitination must be reversed. Dephosphorylation is thought to be accomplished in the endosome by protein phosphatase 2A. Very little is known about the enzymes responsible for deubiquitinating (Millard and Wood, 2006). If one examines the sequence of the C terminus of the A_2A receptor in Figure 3, it is evident that the number of lysine residues to which ubiquitin may be attached is limited. Two of the three lysine residues fall into the region where ARNO and α-actinin are thought to bind. It is evident that the addition of a 76-amino-acid moiety may not be conducive for the interaction with either ARNO or α-actinin given that neither has an ubiquitin-interaction domain.

The presence of multiple binding sites for signalling molecules, scaffolding proteins and adapter proteins may allow the C terminus of the A_2A receptor to serve as a coincidence detector (the binding of agonist and a second signal must occur simultaneously for interactor recruitment or release), as a signal integrator (several inputs must accumulate sequentially before interactor recruitment or release) or as an alternative switch (depending on where the receptor resides or which accessory proteins are available, the receptor may employ different signalling pathways to elicit distinct biological responses). The presence of translin-associated protein X, for instance, allows the A_2A receptor to recruit p53-mediated cell cycle arrest and thus mediate differentiation of PC12 cells in a manner independent of G_s and the cAMP cascade (Sun et al., 2006). At this point, recruitment of ARNO by the A_2A receptor is presumably counterproductive, as ARNO inhibits the hallmark of neuronal differentiation, the sprouting of dendrites and axons (Hernandez-Deviez et al., 2002, 2004).

Glossary

ARF

ADP-ribosylation factor

ARNO

ARF-nucleotide binding site opener

FRET

fluorescence/Foerster resonance energy transfer

GRK

G-protein-coupled receptor kinase

GPCR

G-protein-coupled receptor

MAP kinase

mitogen-activated protein kinase

mGluR5

metabotropic glutamate receptor 5

USP4

ubiquitin-specific protease 4